Acute kidney injury (AKI) is a complex and deadly disease that is strongly associated with progressive loss of kidney function, cardiovascular disease, poor quality of life, and death. The most severe forms of AKI cause parenchymal damage, which manifests as a persistent loss of kidney function. This condition, termed intrinsic AKI (iAKI), carries the highest mortality and risk for long-term loss of kidney function. Due to their older age and high prevalence of risk factors, Veterans are at especially high risk for experiencing iAKI compared to the general population. Despite decades of investment, no successful treatments have been translated from preclinical studies into routine clinical practice. The latter has led to calls for greater understanding of the mechanisms responsible for defining risk in human iAKI. A growing area of investigation is understanding the genetic basis for susceptibility to iAKI. Early studies have been limited by small sample sizes, a lack of unbiased approaches (e.g. genome wide association (GWA)) and predicted gene expression studies, and most importantly, superficial phenotyping that does not distinguish between causes of AKI. As iAKI is a heterogeneous condition, this critical deficiency can dilute biological signals and treatment effects in large-scale studies. Lastly, few studies in AKI have explored identifying novel phenotypes, or endophenotypes of iAKI, which have shown promise for improving understanding of other complex and heterogeneous conditions. The overarching goals of this proposal are to a) advance the clinical phenotyping of the most common and severe forms of intrinsic AKI (iAKI), and b) leverage these phenotypes to identify genetic variants associated with iAKI. In Aim 1, we will apply a data-driven deep learning algorithm to dense structured data and narrative text to discover data patterns that will likely represent a mixture of previously recognized and unrecognized endophenotypes of AKI. In Aim 2, we will complement this strategy by generating probabilistic phenotype algorithms that use manual chart review to identify traditional iAKI phenotypes in 3 clinical settings where iAKI is common: cardiovascular surgery, cardiac catheterization, and sepsis. In Aim 3, we will perform a series of GWA studies within these settings comparing cases identified by our iAKI phenotyping algorithms in Aim 2 to patients without AKI within the Million Veteran Program. We will also conduct the same analyses using the most promising Aim 1 data-driven endophenotypes. We will evaluate the top associated regions using PrediXcan to examine predicted gene expression in kidney tissues. The proposed studies will be performed within the VA Million Veterans Gamma Program by a multidisciplinary team of experts in AKI phenotyping, informatics-based phenotype developers, and genetic epidemiologists. The deliverables from this proposal will advance the computational phenotyping of iAKI, expand the rigor and scale of large-scale genotype-phenotype studies in iAKI, and provide important information regarding clinical iAKI disease mechanisms.